Repository navigation
SIGSEGV on querying using multiple connections #435
Description
What happens?
Concurrently querying select 1 for 8 DuckDBPyConnections, one in each thread, segfaults.
To Reproduce
Minimal reproduction:
import concurrent.futures
import faulthandler
import duckdb
faulthandler.enable()
conns = [duckdb.connect() for _ in range(8)]
def run(i):
conns[i % 8].execute('SELECT 1').fetchall()
with concurrent.futures.ThreadPoolExecutor(8) as ex:
for f in [ex.submit(run, i) for i in range(200)]:
try: f.result()
except Exception as e: print(f'{type(e).__name__}: {e}')
Expected output:
Empty, exit code 0
Actual output
RuntimeError: Could not allocate tuple object!
Fatal Python error: Segmentation fault
System: mac arm64, python3.12, duckdb 1.5.2
Also reproducible on a railway env with python 3.13, linux x86-64, duckdb 1.5.2
OS:
aarch64
DuckDB Version:
1.5.2
DuckDB Client:
1.5.2
Hardware:
No response
Full Name:
David He
Affiliation:
Intercept
Did you include all relevant configuration (e.g., CPU architecture, Linux distribution) to reproduce the issue?
- Yes, I have
Did you include all code required to reproduce the issue?
- Yes, I have
Did you include all relevant data sets for reproducing the issue?
Not applicable - the reproduction does not require a data set
Activity
[-]SIGSEGV on quering parquet file from multiple connections[/-][+]SIGSEGV on querying using multiple connections[/+]evertlammerts commentedon May 18, 2026
Hey @dhe128, thanks for the report. The segfault should definitely not happen, I put up #462 for that.
However, it's not possible to re-use connections across threads. You can create cursors from a connection and use those:
This would work with a single connection as well, btw:
Use recursive mutex to deal with GIL <-> internal lock deadlocks (#462)